Welcoming back my arm: affective touch increases body ownership following right-hemisphere stroke

Abstract Right-hemisphere stroke can impair the ability to recognize one’s contralesional body parts as belonging to one’s self. The study of this so-called ‘disturbed sense of limb ownership’ can provide unique insights into the neurocognitive mechanisms of body ownership. In this study, we address a hypothesis built upon experimental studies on body ownership in healthy volunteers. These studies have shown that affective (pleasant) touch, an interoceptive modality associated with unmyelinated, slow-conducting C-tactile afferents, has a unique role in the sense of body ownership. In this study, we systematically investigated whether affective touch stimulation could increase body ownership in patients with a disturbed sense of limb ownership following right-hemisphere stroke. An initial feasibility study in 16 adult patients with acute stroke enabled us to optimize and calibrate an affective touch protocol to be administered by the bedside. The main experiment, conducted with a different sample of 26 right hemisphere patients, assessed changes in limb ownership elicited following self- (patient) versus other- (experimenter) generated tactile stimulation, using a velocity known to optimally activate C-tactile fibres (i.e. 3 cm/s), and a second velocity that is suboptimal for C-tactile activation (i.e. 18 cm/s). We further examined the specificity and mechanism of observed changes in limb ownership in secondary analyses looking at (i) the influence of perceived intensity and pleasantness of touch, (ii) touch laterality and (iii) level of disturbed sense of limb ownership on ownership change and (iv) changes in unilateral neglect arising from touch. Findings indicated a significant increase in limb ownership following experimenter-administered, C-tactile-optimal touch. Voxel-based lesion-symptom mapping identified damage to the right insula and, more substantially, the right corpus callosum, associated with a failure to increase body ownership following experimenter-administered, affective touch. Our findings suggest that affective touch can increase the sense of body-part ownership following right-hemisphere stroke, potentially due to its unique role in the multisensory integration processes that underlie the sense of body ownership.


Introduction
Several disturbances of body awareness occur after righthemisphere stroke, including asomatognosia (i.e. feelings of non-belonging or non-recognition of the limb) and somatoparaphrenia (delusional ideas of disownership; see Jenkinson et al., 2018). These disturbances reveal that our seemingly effortless sense of body ownership (i.e. the sense that 'my' body belongs to 'me') is actually supported by distinct neurocognitive mechanisms that warrant scientific study. In addition, these symptoms present a significant but unmet clinical challenge, with body unawareness being associated with poor engagement with rehabilitation, longer hospitalization and poorer prognosis (see Jenkinson et al., 2011;Besharati et al., 2014a, b).
Existing research into the mechanisms of body ownership in healthy individuals has identified dynamic integration and weighting processes involving several exteroceptive (e.g. vision) and interoceptive signals (i.e. representing the physiological state of the body; see Craig, 2003). Among these signals, a specific type of pleasant touch (hereafter referred to as affective touch) plays a significant role (Crucianelli et al., 2013(Crucianelli et al., , 2018Lloyd et al., 2013;van Stralen et al., 2014;Ponzo et al., 2018). Affective touch involves slow-conducting, unmyelinated C-tactile afferent (CT afferents) nerve fibres, which are thought to project mainly to the posterior insula via a distinct ascending pathway (Olausson et al., 2002;Mcglone et al., 2012), associated mainly with interoceptive rather than exteroceptive modalities (Craig, 2009;Ceunen and Vlaeyen, 2016). Microneurography studies have shown that CT afferents, located on the hairy skin of the body, respond preferentially to slow (i.e. velocities between 1 and 10 cm/s), dynamic, light-pressure, touch (Vallbo and Hagbarth, 1968;Nordin, 1990;Vallbo et al., 1993Vallbo et al., , 1999. Moreover, their activation is positively correlated with subjective ratings of tactile pleasure in healthy subjects (Lö ken et al., 2009). Importantly, work on peripheral neuropathies, as well as functional neuroimaging studies in healthy participants (see Morrison, 2016 for a review and meta-analysis), supports distinct, yet interacting (Marshall et al., 2019), functional roles between large diameter, fast-conducting, myelinated afferent fibres (Ab-fibres) and the CT-afferent system. For example, patients with a genetic mutation that reduces CTafferent density reported reduced pleasantness in response to CT-optimal touch, while discriminatory aspects of touch remained intact (Morrison et al., 2011). By contrast, a patient with selective loss of large, myelinated afferents showed the reverse pattern, with a faint but preserved sensation of pleasant touch in response to light, CT-optimal stoking touch on the forearm (Olausson et al., 2002). However, recent work in spinal cordotomy patients suggests some overlap in the processing of hedonic and discriminative aspects of touch (Marshall et al., 2019) and further research is needed to specify the precise contribution of CT and non-CT fibres.
Our group was the first to show that applying CT-optimal, affective touch during multisensory illusions of body-part ownership, such as the classic rubber hand illusion (Botvinick and Cohen, 1998), enhances ownership (Crucianelli et al., 2013). Several other studies and groups have now replicated this finding (Lloyd et al., 2013;van Stralen et al., 2014;Panagiotopoulou et al., 2017;Crucianelli et al., 2018;Ponzo et al., 2018) and further investigated different explanations of this effect.
First, it appears that affective congruency between different modalities (e.g. seeing a soft, pleasant material and feeling one via touch) increases feelings of ownership during multisensory integration, over and above other amodal properties such as temporal and spatial congruency (Filippetti et al., 2019). Moreover, CT-optimal, affective touch can reduce sensations of 'deafference' during asynchronous modulation (Panagiotopoulou et al., 2017).
'Deafference' in this context refers to the unpleasant and numb feelings about one's own body caused by the temporal mismatch (a cross-modal property the brain uses to bind sensations together during multisensory integration) between seen and felt tactile stimulation (Longo et al., 2008;Lesur et al., 2019). These feelings of deafference and limb disownership are reported when healthy individuals experience multimodal mismatches (e.g. conflicting visual-proprioceptive-tactile signals) during the asynchronous conditions of multisensory illusions, such as the rubber hand or full body illusion (Longo et al., 2008;Lesur et al., 2019) and mirror box illusion (Medina et al., 2015). Interestingly, patients with peripheral sensory neuropathy experience similar feelings, reporting paraesthesia and numbness in the affected limbs, due to the physical loss of sensory afferent nerve fibres (e.g. Rothwell et al., 1982). Although patients with right-hemisphere stroke with a disturbed sense of limb ownership (DSO; Baier and Karnath, 2008;Jenkinson et al., 2018) typically do not suffer from peripheral neuropathy, they often report similar feeling of deafference, which might arise either from some form of multimodal mismatch, similar to that produced by the temporal asynchrony of multisensory illusions in healthy individuals, or from central sensory deficits. Although to our knowledge no study has compared such groups directly, we hypothesized that CT-optimal, affective touch may reduce deafference feelings and enhance body ownership of the affected arm in right-hemisphere patients with DSO in a similar way as it reduces deafference feelings in healthy individuals.
Moreover, our study was motivated by the fact that previous studies on DSO have only assessed discriminatory and spatial aspects of tactile perception and no study to date has examined whether spontaneous feelings of 'deafference', such as the sensation that the arm is abnormally cold, numb or painful, may contribute to DSO. Indeed, the neuropsychological literature typically refers to somatosensation in relation to tests such as tactile acuity or discrimination, but DSO and impaired somatosensation are known to double dissociate (see Vallar and Ronchi, 2009 for a review). More specifically, although somatosensory deficits are common in patients with DSO, comparable deficits are often present in patients without DSO (see Gandola et al., 2012 for an example) and some DSO patients have no reported somatosensory impairments (see Smit et al., 2019 for a recent example of complete body disownerhship without somatosensory impairment). Moreover, somatosensory deficits can be reduced without a concurrent improvement in disownership (Moro et al., 2004) and vice versa (see Bolognini et al., 2014 for an example of reduced somatoparaphrenia without concurrent changes in other deficits related to the right hemisphere). These findings indicate that tactile, sensory deficits are neither necessary nor sufficient for DSO to occur. Yet it remains unknown whether and how the motivational and affective aspects of somatosensation may relate to DSO. In this study, we focus on one of these modalities, namely the perception of tactile pleasantness, and its relation to body ownership.
Although it can be hard to systematically sample such subjective, spontaneous feelings of deafference by the bedside, the results of a previous study suggested the hypothesis that DSO may be at least partly explained as sensations about the affected arm, which cannot be explained by existing top-down expectations of selfhood (Martinaud et al., 2017). Specifically, we found that almost all of our sample of patients with right perisylvian lesions (N ¼ 31), with and without DSO symptoms, experienced feelings of ownership over a rubber hand within 15 s and without any tactile stimulation (a phenomenon termed visual capture of ownership). Paradoxically, the subset of these patients who had clinical somatoparaphrenia denied the ownership of their own arm, even when they accepted (mistakenly) the ownership of a rubber hand that was placed at the same position on the left hemispace. Thus, DSO is not merely a matter of being unable to attribute a seen left arm positioned in the left hemispace to the self, but rather it is possible that some unexpected sensation or feelings about the arm leads patients to infer that this arm cannot possibly be their own (for the wider theoretical context of this hypothesis, see Fotopoulou, 2015;Crucianelli et al., 2019). Specifically as regards DSO, we proposed (Martinaud et al., 2017) that such 'unexpected' sensations in the arm may be described as feelings of deafference, ultimately leading these patients to feel that the arm they see does not 'feel' as they expect it to feel and hence to infer that the arm is not their own. Unfortunately, in practice, it can be difficult to reliably assess the presence of such spontaneous deafference sensations and their precise role in arm ownership. However, a possible alternative way to test these ideas is to experimentally 'reduce' such sensations using affective touch stimulation (which as aforementioned can attenuate experimentally induced feelings of deafference; Panagiotopoulou et al., 2017) and observe the effect on body-part ownership. Two single-case studies from the same group provide indirect evidence for this proposed effect. First, van Stralen et al. (2011) reported that gentle touch (not controlled for CT-fibre activation) increased feelings of arm belonging and related emotional attitudes in a patient with right-hemisphere stroke with DSO (van Stralen et al., 2011). Second, Smit et al. (2019) report a patient with full body disownership following a tumour resection in the temporoparietal cortex, who experienced an increased sense of ownership towards a rubber hand during slow (CT-optimal) touch. Importantly, neither study provides a direct examination of whether applying affective touch to patients with right-hemisphere stroke can reduce DSO.
A second aspect of touch, which may moderate its effect on body ownership and also requires examination, is whether touch is self-generated or originates from another person (externally generated touch). Traditionally, selfgenerated touch has been associated with attenuation of the resulting tactile stimulation, leading to the wellknown phenomenon that we cannot tickle ourselves (Blakemore et al., 1998a). Recent work in healthy subjects further shows how body ownership influences this sensory attenuation during the rubber hand illusion (Kilteni and Ehrsson, 2017). However, other research highlights the enhancing effect of self-touch on sensory perception (see Valentini et al., 2008;Ackerley et al., 2012), including in patients with right-hemisphere stroke (White et al., 2010) and those with disturbed body ownership (see van Stralen et al., 2011). Attentional modulation and temporal expectation are proposed mechanisms for the observed self-enhancement effect. However, while the perceptual effects and neural basis of self-versus other-generated, CT-optimal, affective touch have been explored by our group (Gentsch et al., 2015) and others (e.g. Boehme et al., 2019), to our knowledge no group study has assessed the potentially beneficial effects that affective, self-versus other-generated touch may have on body ownership in patients with stroke.
On the basis of the above, we aimed to examine how affective touch influences body-part ownership in patients with DSO. We predicted that caress-like touch delivered at CT-optimal velocities, would enhance limb ownership in patients with right-hemisphere stroke with DSO. We also included in our study a CT-suboptimal, fast-touch control condition to establish whether the mechanism underlying any touch-based increase in limb ownership is mediated by the CT system, as opposed to being the result of more generic touch-based, or attentional effects. This faster, but otherwise identical, touch is known to not activate the CT system optimally and is typically judged as more intense but less pleasant than touch activating the CT system (see e.g. Case et al., 2016). We also examined how touch delivered by the patient themselves (self-touch) versus another person (other-touch) might moderate these effects. Because of the opposing effects reported in the extant literature, we did not make a priori predictions regarding which kind of touch (self or other) would produce greater changes in body-part ownership. In secondary, exploratory manipulations and statistical analyses, we aimed to examine the specificity of any touch effects we observed, by investigating the relationship between CT-touch and CT-suboptimal touch perception (intensity and pleasantness ratings), levels of DSO and changes in ownership. We also implemented two control manipulations, in a subset of patients and using only affective touch, to examine the specificity of CTtouch effects. First, we examined the potential effect of affective touch on another common symptom following right-hemisphere stroke, namely unilateral neglect. Second, we assessed whether changes in arm ownership result from a general, positive valence effect, by applying affective touch to the non-affected, ipsilateral (right) arm and measuring changes in ownership of the contralesional (left) arm. Finally, in exploratory lesion analyses, we examined the underlying neuroanatomy of DSO and changes in DSO consequent to touch, which have not been examined in previous studies.

Materials and methods Patients
Forty-two patients with acute, right-hemisphere stroke were recruited by screening consecutive admissions at stroke-rehabilitation wards located in London, UK, and Verona, Italy, as part of larger, ongoing projects on body awareness following right-hemisphere stroke. Inclusion criteria were: (i) right-hemisphere lesion confirmed by clinical neuroimaging (CT or MRI), (ii) dense contralesional hemiplegia (i.e. power < 1; Medical Research Council scale; see assessments below) and (iii) <4 months from symptom onset, (iv) right handed. Exclusion criteria comprised: (i) previous history of neurological or psychiatric illness, (ii) <7 years of education, (iii) medication with significant cognitive or mood side effects and (iv) language impairments that precluded the completion of assessments. Sixteen of these patients were recruited in the UK to an extensive feasibility study, which we conducted prior to the main experiment. In this initial feasibility study, we explored whether patients were able to tolerate a relatively long, touch protocol, perceive the difference between affective and neutral touch and show some difference in their body representation as a result of touch. This study informed our predictions and, importantly, allowed us to develop a robust but also clinically feasible experimental protocol for the main study (full details are reported in Supplementary Materials). The remaining 28 patients were recruited consecutively from both the UK (n ¼ 21) and Italy (n ¼ 7) and took part in only the main experiment. We subsequently excluded two patients from the final data analysis: one due to an experimenter error leading to incorrect administration of the experimental protocol and another containing a large amount of missing data (i.e. completed only one-fourth experimental conditions) due to patient fatigue causing termination of the test protocol; therefore, a total of 26 patients contributed to the final, main analysis (see Table 1 for full details). The study was approved by the relevant University and National Health Service (UK) ethical committees, and patient consent was obtained according to the Declaration of Helsinki.

Clinical and cognitive assessment
Patients underwent a neurological and neuropsychological examination comprising: a clinical assessment of left-right orientation (during which patients were asked to identify visually or use their unaffected side to point to/touch their own/the experimenter's left/right hand, ear and leg), assessment of motor ability (power) of the left upper and lower limbs (Medical Research Council scale; Guarantors of Brain, 1986), upper and lower visual fields and tactile extinction on the upper and lower limbs (using the Bisiach et al., 1986 technique), light touch perception on the left arm (revised Nottingham Sensory Assessment; Lincoln et al., 1998), orientation in time and space as well as general cognitive ability (Montreal Cognitive Assessment; Nasreddine et al., 2005), estimated premorbid intelligence (Weschler Test of Adult Reading; Wechsler, 2001), working memory (using the verbal digit span subtest of the Wechsler Adult Intelligence Scale III; Wechsler, 1997), visuospatial neglect [using five subtests (star cancellation, line bisection, line crossing, copy and clock drawing) of the Behavioural Inattention Test; Halligan et al., 1991], personal neglect (one-item test; Bisiach et al., 1986;and comb/razor test;McIntosh et al., 2000), anosognosia for hemiplegia (Bisiach et al., 1986;Feinberg et al., 2000), executive function (frontal assessment battery; Dubois et al., 2000) and brief assessment of depression and anxiety (Hospital Anxiety and Depression Scale; Zigmond and Snaith, 1983).

Experimental design and measures
To examine the effect of touch on limb disownership, we employed a two (touch valence: affective versus neutral) Â two (touch administrator: self versus other) repeatedmeasures design. The valence of touch (affective versus neutral) was manipulated using two different velocities, such that touch was administered using either a slow, CT-optimal (3 cm/s) or fast, CT-suboptimal (18 cm/s) touch, which are typically rated as relatively high versus low in terms of tactile pleasantness and can be reliably distinguished from each other (see Crucianelli et al., 2013Crucianelli et al., , 2018Ponzo et al., 2018). This touch was administered by either the patient (self) or experimenter (other) in a block design (see 'Procedures' section), whereby four affective or neutral touch trials were delivered per block (plus one sham trial in which the movement was simulated but no touch was delivered to the forearm). The order of conditions was only partly counterbalanced, based on what our sample size and practical, bedside considerations allowed; thus, we examined the possible influence of these order effects in specific analyses, below.
The primary dependent variable was 'ownership change'. The level of disownership during the experimental session was assessed prior to any touch conditions (i.e. baseline disownership) and after each touch condition (i.e. post self-neutral, other-neutral, self-affective and other-affective touch) by asking three questions designed to assess arm recognition, feelings of belonging and existence (as discussed by Jenkinson et al., 2018): (Q1) 'Is this (pointing to the patient's left arm) your own arm?', (Q2) 'Does it ever feel like this (patient's left) arm does not belong to you/is not really yours?' and (Q3) 'Do you ever feel that your left arm is missing/has disappeared?'. To avoid repetition and loss of attention, the exact wording used was varied on a trial-by-trial basis as indicated Orientation U (person, time and place) Digit span forwards (maximum number repeated) BIT-line bisection (maximum  above (see Supplementary Materials for further details of the feasibility study used to develop this procedure). Patient responses to each question were recorded verbatim and scored by the experimenter using an existing scoring procedure (Feinberg et al., 2000 method adapted to assess disownership rather than awareness), where higher scores indicated greater levels of disownership: 0 ¼ no disownership; 0.5 ¼ partial disownership; and 1 ¼ disownership (see also Martinaud et al., 2017 for a further example of this method). This method typically invites a binary response to the closed, direct questions asked and also allows for some ambiguity and comments made by the patient to be captured and scored. Specifically, a response that unambiguously indicated a lack of limb recognition, belonging or existence (i.e. answering 'no' to Q1 or 'yes' to Q2 or Q3) was scored 1, while a response that indicated intact limb recognition, belonging or existence (i.e. answering 'yes' to Q1 or 'no' to Q2 or Q3) was scored 0. Any response to these three questions that indicated uncertainty regarding limb recognition, belonging or existence (e.g. answering 'not sure', 'I don't know' or giving some other ambiguous response) was scored 0.5. A second, independent researcher (P.M.J.) read the verbatim-recorded response sheets and verified the initial experimenter score, which primarily involved assessing whether any ambiguous answers should indeed be scored 0.5 or given a binary score. Any disagreement between these two ratings was resolved by discussing and agreeing the score with a third, senior researcher (A.F.) who was blind to experimental condition. To calculate 'ownership change', we totalled the raw scores (total disownership score minimum ¼ 0, maximum ¼ 3) and then subtracted the post-touch disownership score from the pre-touch baseline (i.e. baseline score À post-touch score), so that a post-touch increase in ownership was indexed by a positive ownership change score, while any increase in disownership was indexed by a negative ownership change.
To explore the specificity and mechanism of the CTtouch effect, we also conducted a number of control tasks in a subset of patients and trials (constrained by fatigue and other practical considerations, these were merely exploratory investigations; see Supplementary Materials for details), to examine (i) the effect of touch on visuospatial and personal neglect and awareness of neglect and (ii) whether touch applied to the 'right' arm would influence disownership of the 'left' limb, and the effect of CT and CT-suboptimal touch perception on ownership change.

Procedures
Touch was administered to the dorsal surface of the forearm (in a preselected area between the wrist and elbow crease; $18 cm), by a female experimenter using a soft make-up brush made from natural hair (Natural hair Blush Brush, No. 7, The Boots Company). Visual feedback of the touched arm was blocked either by asking the patient to wear a blindfold or (if patients refused to wear a blindfold) by closing their eyes while an assistant experimenter also held a cardboard carton over the touched arm to obscure a direct view. Each touch 'trial' comprised four strokes in alternating directions, from the elbow to the wrist and vice versa, with a pause of 1 s between each stroke to reduce fibre fatigue. Slow touch strokes at a velocity of 3 cm/s were administered in 6 s over the 18-cm-long area, and fast strokes over the same area lasted 1 s. Each 'block' comprised four trials of the same prespecified touch velocity (i.e. 4 Â 4 touch trials ¼ 16 strokes per block), and one randomly inserted sham trial, during which the brush was held several centimetres above the touch location and moved to mimic the slow/fast touch, but did not make contact with the patient's skin (see Fig. 1 for a schematic representation of the experiment).
During a pre-experimental phase, patients were introduced to the stroking apparatus, procedure (i.e. slow, fast and sham trials) and vertical response scale used to assess the perceived intensity and pleasantness of touch (see Supplementary Materials for details). A vertical scale was used to minimize the effects of neglect; we also always ensured that the participants could see the scale and read it aloud to facilitate them). Understanding of the response scale was also established by asking patients to rate the pleasantness of various physical and conceptual items (i.e. to be touched by a thorn/cotton on the skin, to win the lottery, to lose your keys, to be praised by someone you love), with any misunderstandings clarified prior to continuing to the main experimental protocol. One trial of slow/fast touch was also administered during this preexperimental stage to establish if the patient could perceive touch on the left arm.
During 'other-touch' blocks, the experimenter held the brush and touched the patient in the required location (left or right arm) using the appropriate velocity (slow or fast). By contrast, during 'self-touch' blocks, the patient was physically guided (by the experimenter holding/supporting the patient's hand/arm and brush when necessary) to touch the relevant location on the contralateral arm using the appropriate speed (for this method of selftouch, see White et al., 2010). This physical support was provided during all self-touch trials [i.e. both when the unaffected (right) and hemiplegic (left) arms were the 'active' arms delivering the touch] to ensure consistency between self and other stroking in terms of velocity and pressure of stroking. Immediately after each trial (four strokes), the patient was asked to rate the touch intensity and pleasantness. Ownership was assessed prior to any touch (i.e. pre-touch baseline) and again following each touch block. Neglect was assessed before and after each block of affective touch only, due to time and fatigue constraints. A 1-min break was introduced between blocks, with longer breaks given where necessary (i.e. patient appeared fatigued or requested a break).

Statistical analyses Main experimental measure-ownership change
Our main analysis aimed to determine whether and to what extent different types of touch would increase body ownership; therefore, only patients with some baseline level of disownership on the day of the experimental session (i.e. baseline DSO score !0.5; DSOþ patients; n ¼ 20) were suitable for the main analysis. Patients not expressing any DSO during the baseline assessment on that day (DSO score ¼ 0; DSOÀ) were not appropriate for this analysis since they were asymptomatic for the primary behaviour of interest (i.e. they did not show disownership during the session) and were thus excluded from it; however, these patients provided relevant comparison data for feasibility, comprehension, suggestibility and other practical considerations applying to our experimental procedures in an acute stroke setting, as well as clinical and neuropsychological measures, lesion analysis and experimental control conditions (described in Supplementary Materials).
We used both frequentist and Bayesian statistics to assess the observed effects, depending on the aim and hypothesis in each case. The complementary use of these two statistical approaches is recommended by a number of authors to facilitate a fuller understanding of the data (see e.g. Howard et al., 2000;Dienes, 2014;Dienes and Mclatchie, 2018;Quintana and Williams, 2018). For frequentist statistical inference, we assessed normality via visual inspection of histograms and the Shapiro-Wilks test. Two patients did not complete one of the four experimental conditions. We therefore analysed the experimental data using both pairwise (i.e. case-by-case) and listwise deletion methods for dealing with missing data, to assess the impact of this missing data on the experimental findings. Bayesian statistics were performed to allow further interpretation of the observed effects, in particular, the extent to which data provided support for the alternative versus null hypotheses. Bayes factors (BF 10 ) indicate the relative strength for the alternative versus null hypotheses (i.e. the number of times more likely the data are under the alternative than the null hypothesis) and were used as a means of interpreting evidence for each hypothesis, using benchmarks provided by Jeffreys (1961). We interpreted a BF 10 of >3 as substantial evidence for the alternative hypothesis, a BF 10 of <0.3 as substantial evidence in favour of the null hypothesis and a BF 10 of >0.3 < 3.0 as insensitive, weak or anecdotal evidence for either hypothesis (see Dienes, 2014;Quintana and Williams, 2018).
We examined the extent to which touch modified body ownership using parameter estimates [95% confidence (frequentist)/credibility (Bayesian) intervals], and Figure 1 Schematic representation of the study procedure. Patients first completed a pre-touch baseline measure of ownership, followed by a block of touch (self-affective, other-affective, self-neutral or other-neutral), and a post-touch measure of arm ownership. A sham (no-touch) trial (not shown above for simplicity) was randomly inserted into each block. Neglect was assessed in a subsample of patients and only pre-post affective touch conditions. Four main blocks completed in total (self-affective, other-affective, self-neutral, otherneutral). Right arm control touch blocks were administered following the same method (without neglect assessment) in a subset of patients.
calculating non-parametric/Bayesian, one-sample t-tests (owing to the limited sample size and non-normal distribution; Shapiro-Wilks test), testing the null hypothesis of zero change in body ownership as a result of the touch applied (i.e. one-sample test value ¼ 0). We corrected for multiple comparisons using a Bonferroni-corrected alpha of 0.0125 (i.e. 0.05/4; applicable to frequentist statistics only). Although we expected touch to increase ownership in patients with DSOþ, we could not rule out the possibility that touch might reduce body ownership, and so we performed all statistics using two-tailed tests. All behavioural analyses were performed using JASP (JASP Team, 2019). Figures for behavioural data were generated in R (R Core Team, 2013) using ggplot 2 (Wickham, 2016) and in MRICron (Rorden and Brett, 2000;Rorden et al., 2007) for lesion analysis.

Clinical and neuropsychological variables
Clinical and neuropsychological data were summarized by grouping patients according to their score on the experimental measure of DSO taken during the pre-touch baseline of the experimental session. As is typical in neuropsychological research with patients with acute stroke, not all assessments were fully completed by all patients, owing to discharge, scheduling issues or patients becoming too ill to complete assessments. In these cases, data are included where available and indicated where missing. Differences in the clinical and neuropsychological variables between these groups were analysed for exploration only, given that these groupings comprised small patient numbers and were not used as the basis for our main experimental manipulation. We used non-parametric tests (Kruskal-Wallis test), and an alpha of 0.01 to account for false positives arising from multiple comparisons, while also avoiding being overly conservative.

Control analyses
A number of control tasks were conducted in a subset of our patients due to practical constraints (as detailed above and in Supplementary Materials) to look at the possible influence of order, unilateral neglect, perceived pleasantness and intensity of CT and CT-suboptimal touch and touch lateralization on body ownership change. There were no order effects and no significant relationship between touch pleasantness or intensity and baseline levels of body ownership or body ownership change. Moreover, our touch manipulation had no effect on neglect, nor did a control application of the touch protocol to the right arm affect the ownership of the affected, left arm in our patients, confirming the specificity of our main findings. Thus, as none of these factors were found to relate to ownership changes in our study (our main measure), the results of these control analyses are presented in Supplementary Materials only.

Lesion analyses
Univariate voxel-based lesion-symptom mapping (Bates et al., 2003;Rorden et al., 2007) was used to identify anatomical regions associated with (i) disownership scores obtained during the baseline (pre-touch) condition of the experimental task (n ¼ 24) and (ii) ownership change scores obtained from the other-pleasant touch condition (n ¼ 18). These voxel-based lesion-symptom mapping analyses, when strictly corrected for multiple comparisons and minimum statistical power, did not show any significant results in our sample, possibly due to the small sample sizes, large lesions and the fact that our experiment was optimized for subjective ownership changes in acute patients at the behavioural level (hence our main dependent variable had a relatively small variance range). In exploratory analyses, we, therefore, repeated these analyses with less restrictive criteria (with no minimum lesion overlap and using 1% false discovery rate correction for multiple comparisons). We report in the main results below only the significant findings from the exploratory lesion analyses. Full details of the methods and results for these lesion analyses are reported in the Supplementary Materials.

Data availability
The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Clinical and neuropsychological results
The clinical and neuropsychological characteristics of patients are summarized in Table 1. Patients did not differ significantly in their clinical or neuropsychological profile. Importantly, even though the revised Nottingham Sensory Assessment of somatosensory function indicated that the perception of light touch was impaired in a number of patients, their ability to perceive both the pleasantness and intensity of repeated dynamic touch to some degree was confirmed in our control task that included sham trials to control for tactile perception (see Supplementary Materials).

Ownership change following touch
There was considerable variability in the extent to which each type of touch led to a change in ownership (see Fig. 2). Despite this variability, a series of one-sample Wilcoxon signed-rank tests (with Bonferroni-corrected a ¼ 0.125 and pairwise deletion of cases with missing data) indicated that body ownership increased significantly following other-affective touch [V(18) ¼ 71, P ¼ 0.012] but not following other types of touch [self-affective: V(20) ¼ 53.50, P ¼ 0.268; self-neutral: V(20) ¼ 67, Fig. 2]. This analysis was also run using listwise deletion for missing data and Bayesian statistics (full details given in Supplementary Materials), confirming a substantive increase in body ownership following other-affective touch (BF 10 ¼ 6.95) but not other types of touch (all other BF 10 > 0.3 < 3.0). In addition to the formal assessment of ownership change reported above, during post-experiment debriefing, we discussed any changes in body ownership with patients clinically, to understand the nature of their experience. One patient, who used to call her left arm her 'alienated arm' in the first days following her stroke, told us that, after the affective touch experiment, she would use her right, intact arm to stroke her left arm and speak to it; she said 'come, correspond to me'. Another patient, somewhat similarly, told us a few days after the experiment: 'I woke up and I called this (left) arm "a beast". It was not my arm, I did not want it, it was some foreign fellow. But then you touched it and I caress it as you said and I decided to love it again. I said "Come, I accept you. I welcome you back" . . . We have been through a lot together.'

Lesion analyses
Our first exploratory lesion analysis, examining brain damage associated with baseline levels of DSO during the experimental session, did not reveal any significant results (reported in Supplementary Materials only, for brevity). The second analysis, of lesions associated with a failure to increase limb ownership following other-affective touch, indicated that damage to the right insula, and more substantially the right corpus callosum, was associated with a failure for body ownership to increase following other-affective touch (false discovery rate-corrected P < 0.01 for z > 2.42; see Fig. 3; full details of both analyses are provided in Supplementary Materials).

Discussion
This study shows for the first time that receiving CToptimal, affective touch can increase the sense of arm ownership in patients with right-hemisphere stroke with DSO. These results are consistent with a growing body of experimental research in healthy subjects, which shows that affective, CT-optimal touch enhances feelings of body-part ownership during multisensory body ownership illusions (Crucianelli et al., 2013(Crucianelli et al., , 2018Lloyd et al., 2013;van Stralen et al., 2014;Panagiotopoulou et al., 2017;Ponzo et al., 2018). Below, we consider the mechanisms by which CT-optimal touch administered by another person increases body ownership.
First, affective touch may reduce the unusual, sensory symptoms often reported by patients with right-hemisphere stroke, in particular 'feelings of deafference'. We have previously argued that DSO may be the result of an inability to integrate current sensory signals with prior expectations about the sensory state of the body (Martinaud et al., 2017). Hence, disownership may occur when patients are unable to use current, aberrant signals from the body (e.g. feelings of 'heaviness', 'numbness', 'coldness' and other similar sensations) to update predictions and expectations about how the affected body parts should feel (Besharati and Fotopoulou, 2020). This explanation is consistent with predictive coding accounts of body ownership, which propose that body ownership involves top-down, cross-modal predictions about the most likely cause of one's bodily experience (e.g. the self) and requires the integration of exteroceptive, proprioceptive and interoceptive sensory information (see Seth, 2013;Apps and Tsakiris, 2014;Fotopoulou, 2015 for further details and discussion). Indeed, research in healthy subjects using the rubber hand illusion shows that a temporal mismatch between seen and felt tactile stimulation (i.e. asynchronous, unpredictable touch) can produce several, unusual 'deafference' sensations similar to those commonly reported by patients with DSO, including pinsand-needs, numbness and the hand being experienced as less vivid (Longo et al., 2008;Lesur et al., 2019). CT-optimal, affective touch, which provides an additional, cross-modal means of integrating different signals across modalities (namely, 'affective congruency'; see Filippetti et al., 2019 for further discussion) and has been found to reduce 'deafference' in healthy subjects (see Damaged MNI voxels predicting failure to increase body ownership following other-affective touch (P < 0.01 for z > 2.42 FDR corrected). (C) Heat map (AUROC) of the voxels with enough power to detect a significant result ranging between 0.5 (minimum power; shown in blue-black) and 1 (maximum discrimination power; shown in red). (D) Three-dimensional reproduction of lesions superimposed onto the Natbrainlab atlas (left) and MNI template (right). FDR ¼ false discovery rate; MNI ¼ Montreal Neurological Institute; VLSM ¼ voxel-based lesion-symptom mapping. Panagiotopoulou et al., 2017), may similarly lessen feelings of deafference in patients with DSO.
Affective touch may also increase arm ownership by enhancing interoceptive signals known to play an important role in the 'feeling of mineness' (see Craig, 2003;Crucianelli et al., 2018). Interoception informs the mind about how the body is doing in relation to certain homeostatic needs (e.g. hunger, thirst, pain and pleasure) and is considered by some modern neuroscientific theories to be the basis of subjective feeling states and the sentient self (Craig, 2009;Seth, 2013;Fotopoulou and Tsakiris, 2017). Thus, in the present study, CT-optimal touch may change beliefs about body ownership by strengthening these fundamental feelings of mineness. This effect is specific to the touched limb, and not a general valence effect, as demonstrated by the fact that applying affective touch to the unaffected right arm did not increase ownership of the affected left arm.
Finally, touch from another person may draw attention to the left hand, thereby enhancing the salience (i.e. the epistemic importance of the hand in relation to other percepts) of signals arising from the affected limb. An increase in attention might result in signals from the affected limb being given greater weight during multisensory integration-which can produce an update in knowledge about the state of the body and change in beliefs about limb ownership (see e.g. Bolognini et al., 2014;D'Imperio et al., 2017). In the context of this explanation, touch from another person might be deemed more salient owing to its lack of concurrent efferent information and consequently greater unpredictability (Blakemore et al., 1998b(Blakemore et al., , 1999. However, our results suggest that enhanced salience or attention is unlikely to be the main mechanism of change in our patients, since visuospatial and personal neglect for the affected side and arm were not consistently or concurrently increased by affective touch. The observed effect of CT-optimal touch appears to be specific to arm ownership; however, the results of our neglect control conditions only allowed for qualitative evaluation of the affective touch effects. Further research is needed to establish with greater certainty the specificity of this effect. Our supplementary, exploratory analyses also indicated that not all patients were able to perceive touch reliably, or the difference between affective and neutral touch intensity and pleasantness (as indicated by our analysis of touch intensity and pleasantness reported in Supplementary Materials). There was also no relationship between the perceived intensity or pleasantness of touch and baseline disownership or the ownership change. Earlier reports found that self-touch enhances the perception of touch in a small group of right-hemisphere patients (White et al., 2010). Our own exploratory examination of the perceived intensity and pleasantness of touch revealed that the 'intensity' of self-touch was greater than that of other-touch (consistent with White et al.), irrespective of velocity, while the perceived 'pleasantness' of touch did not differ for self-versus other-touch or different velocities. Thus, our findings show that even when CT-optimal touch received from an 'other' is perceived as less intense and as equally pleasant as self-touch, it can be particularly potent to increase limb ownership. This finding supports the observed dissociation between neural systems responsible for discriminative versus affective aspects of touch (see McGlone et al., 2014 for a review) and suggests that this dissociation includes the selective enhancement of discriminative self-touch, while not affecting perception of the CT-touch system. However, these findings need to be interpreted with caution as we had limited measurements of patients' tactile perception, either by standardized assessments or, during the task (only four trials per touch type). Future studies should explore CT touch effects in association with systematic, somatosensory assessments.
Our study also has important implications for the rehabilitation of patients with typically poor therapy adherence and engagement. A simple-to-administer, non-verbal, non-invasive affective touch procedure requiring no specialist training or equipment can provide an effective means of improving post-stroke DSO. Despite our averaged findings and the consistent spontaneous remarks from some of our patients, we note that not all patients with DSO showed increased ownership following otheraffective touch and we did not include follow-ups in the present, experimental study. More generally, further research is needed to identify minimum, critical factors that determine who can benefit from affective touch and to establish the optimum dose (frequency, timing and duration of treatment), extent and stability of effects. In addition, further work is needed to explore factors that might moderate the efficacy of affective touch, such as the ability to perceive touch and affective touch, baseline degree of disownership and concurrent neuropsychological deficits.
One potentially fruitful avenue to explore is the identification of neuroanatomical predictors (Lunven et al., 2015;Forkel and Catani, 2018). Our initial lesion analyses did not yield any significant results, likely due to the small sample, large lesions and limited range of scores possible on our behavioural measure. However, our exploratory voxel-based lesion-symptom mapping using less restrictive criteria revealed that lesions in the insula and corpus callosum resulted in significantly less increase in body ownership following other-affective touch. These findings are consistent with the proposed role of the insula in body awareness, as well as the importance of white matter tracts in self-awareness (Pacella et al., 2019) and body ownership (Moro et al., 2016;Martinaud et al., 2017), and warrant further study.
In conclusion, we report novel findings from patients with right-hemisphere stroke with disturbances in body ownership. Using simple tactile stimulation parameters that are known to activate the CT system optimally, we were able to increase the sense of arm ownership in patients with DSO. Based on a number of additional manipulation, we suggest that these interoceptive signals bring about change in ownership by reducing feelings of deafference and allowing new sensations from the affected body parts to be integrated with one's multimodal, self-representation, rather than less likely mechanisms of attentional enhancement. Finally, our study provides the first experimental evidence for the use of neurophysiologically specified type of affective touch in the treatment of DSO that can be further tested in translational studies.

Supplementary Material
Supplementary material is available at Brain Communications online.